一张表,多种查询:跨生成、插补与预测的通用表格模型基准测试

Media reports Hugging Face Papers Source language: Chinese

论文提出 One Table Benchmark,覆盖 56 个数据集、46 个注册方法和 20 个能力维度,在对齐的数据划分下统一评估表格生成、插补与预测三项任务,发现三者共享有用的条件信息,但分布保真度的模型排名与其他任务不同甚至相反。

Full text

A large aligned benchmark reveals that tabular generation, imputation, and prediction share useful conditional information but show distinct fidelity rankings and objective interactions, motivating unified models that improve categorical reconstruction and joint sampling.

Sources

← Back to Daily AI Report